The Market Cheered. America Didn't.

The numbers stopped agreeing with each other somewhere in 2024. As of 22 May 2026, AI-linked equities have outperformed the non-AI S&P 500 by 121 percentage points since the start of 2024, according to Polymarket data. In the same window, U.S. consumer sentiment has crashed to a record low, with 57% of consumers telling surveyors that high prices are actively eroding their personal finances. The market is pricing a productivity revolution. The grocery shopper is not.

That divergence is the story. Not the AI trade itself — that's been told and re-told by every financial wire from here to Singapore — but the gap between what the market is doing and what the people the market is supposed to represent are actually experiencing. One of those data sets is voting in a market. The other is voting in an election. Only one of them is being treated as the reliable signal.

The Numbers Don't Lie — But They Don't Tell the Whole Story Either

The 121-percentage-point spread is structurally significant. It means that investors who loaded up on AI-adjacent names — chip designers, cloud infrastructure providers, robotics firms, power management companies — have seen returns that bear almost no relationship to the broader economy those companies ostensibly serve. A 121-point spread over roughly two and a half years is not noise. It is a signal that the market has decided AI is a discontinuous event, a step-change in corporate profitability that has already arrived, not one that might arrive.

The consumer sentiment data tells a different story. Record lows are not recession-speak for "things are slowing down." They are recession-speak for "the distribution of gains has gone wrong." Fifty-seven percent of consumers saying high prices are eroding personal finances is a statement about purchasing power, not a statement about inflation expectations. It says that the income side of household ledgers is not keeping pace with the cost side — and that this has been true long enough for the language to harden into a settled conviction rather than a passing anxiety.

The two data sets are not measuring the same thing. One is a forward-looking market valuation. The other is a backward-looking survey of how people are actually living. That they move independently is not new. What is new is the scale of the divergence and the explicitness with which it maps onto a single technology sector.

Whose Rally Is It Anyway?

AI stock ownership is not evenly distributed. By most estimates, the top decile of American households by wealth holds roughly 80 to 90 percent of equity market value directly or through defined-contribution pension vehicles. The ownership structure means that a 121-point outperformance in a specific sector accrues overwhelmingly to people who were already relatively well-positioned before the trade began. This is not an observation about intent. It is an observation about who the market's productivity gains are actually reaching.

The workers whose productivity AI is most plausibly augmenting — warehouse associates, call center staff, entry-level clerical workers, gig economy drivers — do not hold significant equity positions. They are consumers first. When their wages are not keeping pace with costs, they are the 57 percent. When AI-driven efficiency gains flow primarily to margins, retained earnings, and share buybacks, the gains do not circulate back through wages at anything like the rate the productivity narrative implies. The market is rewarding capital. The consumers are waiting for a raise that the data suggests is not coming.

The Distributional Machine

The structural frame here is not complicated, but it is routinely obscured by the language the financial press uses to cover equity markets. "AI stocks rally" treats the rise in share prices as the story. The distributional consequence — that this rally transfers purchasing power from no-one-in-particular to people who already hold assets — is treated as background noise.

The same dynamic plays out in reverse for the consumer sentiment figures. "Consumer sentiment falls" treats the survey response as the story. The structural consequence — that households are absorbing cost increases that their income cannot offset — is often framed as a confidence problem, a psychological state that might lift with better news. It is not presented as what it is: a material condition in which the gains from economic activity are flowing upward faster than the costs are being shared.

The market, in this framing, is not a neutral aggregator of information about the real economy. It is a legal and institutional architecture that determines who gets paid and who pays. When it and the survey data say different things, the honest move is to ask which instrument is doing more work — and whose interests that instrument serves.

What Comes Next

The 35% Polymarket probability assigned to Japan following the U.S. in releasing UFO files is the kind of data point that keeps the internet occupied. It is not the kind of data point that changes the distributional arithmetic of American household economics. The stakes of the AI-sentiment divergence are more immediate and more tractable.

If the current configuration holds — AI equities pricing a revolution that hasn't reached household purchasing power — the political pressure will intensify. Sentiment surveys like the University of Michigan's consumer confidence index are not academic curiosities. They feed into consumer spending forecasts, which feed into corporate revenue guidance, which feeds back into equity valuations. At some point the real economy stops cosplaying the market's optimism. The question is whether that reckoning arrives before the structural drivers of the divergence change, or after.

The 121-point spread is a warning about what happens when financial markets price a technology before its distributional effects have materialized. The 57 percent figure is evidence that those effects have not yet materialized for most people. The distance between those two facts is the distance between a rally and an economy. Closing it requires either a wage shock — real income growth strong enough to catch households up — or a market correction. History suggests the latter happens faster than the former.

This article was filed from wire and Polymarket data as of 22 May 2026. Monexus covered the AI equity outperformance as a market story; the consumer sentiment data was covered as a polling story. We are proposing to run them together.